Research And Development Of Fatigue Driving Detection System Based On Machine Vision

In recent years,with the rapid development of China’s economy,the number of cars has increased year by year,and the frequency of road traffic accidents has also increased.The research shows that most traffic accidents are mainly caused by improper operation of drivers,among which the casualties and property losses The traditional fatigue driving detection method makes judgment by comparing the driver’s normal driving state with the vehicle driving information,but it is easy to be affected by road conditions,and the detection accuracy is low.Or collect the physiological characteristics of the driver for judgment,but the cost is too high,and the detection equipment worn will affect the normal driving of the driver.Fatigue driving detection based on machine vision method,can obtain higher detection accuracy at the same time will not affect the driver’s normal driving,so this article use machine vision related theory,by building a lightweight convolution neural network for face detection,and then get driver’s eye and the change of the fatigue characteristics of the mouth,judgment and warning for fatigue driving.The designed fatigue driving detection system can detect fatigue driving quickly and accurately.The main work of this paper is as follows:1.Replace the backbone feature extraction network in the YOLOv4 target detection algorithm with the mobilenetv3 lightweight feature extraction network to build a mobilenet-YOLO lightweight face detection algorithm to ensure the accuracy of face detection and increase the speed of face detection.2.Use the facial feature point detection model in the dlib library to obtain the coordinates of the driver’s 68 facial feature points,and calculate the driver’s eye(EAR)to mouth(MAR)aspect ratio from this,and perform PERCLOS eye fatigue The feature judgment and the mouth fatigue feature judgment of the yawn behavior detection are used to obtain the fatigue judgment of the fatigue driving detection system.3.Using Jetson nano 2G deep learning embedded platform as the main control core of the fatigue driving detection system,image acquisition is carried out through the camera,and the buzzer module is used to remind the fatigue driving to build a machine vision fatigue driving detection system.The experimental results show that through the Jetson nano 2G deep learning embedded platform,the fatigue driving test set is used to test the face detection algorithm,the detection accuracy rate is 98.17%,the average detection speed is 5.69frames/s;Through PERCLOS eye fatigue characteristic judgment standard test,the eye fatigue threshold of the eye EAR is determined to be 0.97;in the test process of the Yawdd yawn data set,the yawn threshold of the mouth MAR is determined to be 0.32;Through the verification of the fatigue driving detection system carried in the real vehicle,the system can operate normally and give effective early warning to drivers in fatigue state.